Ready-to-use
GEMINI.mdand.gemini/settings.jsonfor your project. Clone, copy, edit.
Part of agent-anatomy — boilerplates for every AI coding agent.
git clone https://github.com/agent-anatomy/gemini.git .gemini-boilerplate
cp .gemini-boilerplate/GEMINI.md ./GEMINI.md
cp -r .gemini-boilerplate/.gemini ./.gemini
rm -rf .gemini-boilerplateThen edit GEMINI.md to match your project.
GEMINI.md ← project instructions, always read
.gemini/
└── settings.json ← Gemini CLI configuration
GEMINI.md is the project instruction file for Gemini CLI (Google's AI coding agent). Gemini reads it automatically at the start of every session — no configuration needed.
Equivalent to CLAUDE.md for Claude Code, or AGENTS.md for Codex.
What to put in GEMINI.md:
- Project description and stack
- Build, test, and lint commands
- Coding conventions
- What Gemini should never touch
- Architecture context
Like Claude Code, Gemini CLI supports GEMINI.md in subdirectories. Root file always loads. Subdirectory files load when Gemini works in that subtree.
Use for monorepos where different packages have different conventions.
| File | Commit? |
|---|---|
GEMINI.md |
✅ Yes |
.gemini/settings.json |
✅ Yes |
Does Gemini read GEMINI.md automatically? Yes. Every session, no setup required.
Can I have GEMINI.md in subdirectories? Yes. Gemini merges root + subdirectory instructions when working in that subtree.
What goes in .gemini/settings.json? Gemini CLI configuration — auth type, theme, and other preferences.
Is GEMINI.md the same concept as CLAUDE.md? Yes. Both are session instruction files committed to git so the whole team benefits. Different agent, same idea.
| Agent | Boilerplate |
|---|---|
| Claude Code | agent-anatomy/claude |
| Cursor | agent-anatomy/cursor |
| OpenAI Codex | agent-anatomy/codex |
| Windsurf | agent-anatomy/windsurf |
| GitHub Copilot | agent-anatomy/copilot |
| Aider | agent-anatomy/aider |